Point Cloud Unsupervised Pre-training via 3D Gaussian Splatting

• URL: http://arxiv.org/abs/2411.18667v1
• Date: Wed, 27 Nov 2024 16:11:45 GMT
• Title: Point Cloud Unsupervised Pre-training via 3D Gaussian Splatting
• Authors: Hao Liu, Minglin Chen, Yanni Ma, Haihong Xiao, Ying He,
• Abstract summary: We propose an efficient framework named GS$3$ to learn point cloud representation. Specifically, we back-project the input RGB-D images into 3D space and use a point cloud encoder to extract point-wise features.
• Score: 7.070581940661794
• License:
• Abstract: Pre-training on large-scale unlabeled datasets contribute to the model achieving powerful performance on 3D vision tasks, especially when annotations are limited. However, existing rendering-based self-supervised frameworks are computationally demanding and memory-intensive during pre-training due to the inherent nature of volume rendering. In this paper, we propose an efficient framework named GS$^3$ to learn point cloud representation, which seamlessly integrates fast 3D Gaussian Splatting into the rendering-based framework. The core idea behind our framework is to pre-train the point cloud encoder by comparing rendered RGB images with real RGB images, as only Gaussian points enriched with learned rich geometric and appearance information can produce high-quality renderings. Specifically, we back-project the input RGB-D images into 3D space and use a point cloud encoder to extract point-wise features. Then, we predict 3D Gaussian points of the scene from the learned point cloud features and uses a tile-based rasterizer for image rendering. Finally, the pre-trained point cloud encoder can be fine-tuned to adapt to various downstream 3D tasks, including high-level perception tasks such as 3D segmentation and detection, as well as low-level tasks such as 3D scene reconstruction. Extensive experiments on downstream tasks demonstrate the strong transferability of the pre-trained point cloud encoder and the effectiveness of our self-supervised learning framework. In addition, our GS$^3$ framework is highly efficient, achieving approximately 9$\times$ pre-training speedup and less than 0.25$\times$ memory cost compared to the previous rendering-based framework Ponder.

Related papers

• SimC3D: A Simple Contrastive 3D Pretraining Framework Using RGB Images [42.69443644770913]
  SimC3D is a 3D contrastive learning framework for pretraining backbones from pure RGB image data. Traditional multi-modal frameworks facilitate 3D pretraining with 2D priors by utilizing an additional 2D backbone. SimC3D directly employs 2D positional embeddings as a stronger contrastive objective, eliminating the necessity for 2D backbones.
  arXiv  Detail & Related papers  (2024-12-06T18:59:04Z)
• PFGS: High Fidelity Point Cloud Rendering via Feature Splatting [5.866747029417274]
  We propose a novel framework to render high-quality images from sparse points. This method first attempts to bridge the 3D Gaussian Splatting and point cloud rendering. Experiments on different benchmarks show the superiority of our method in terms of rendering qualities and the necessities of our main components.
  arXiv  Detail & Related papers  (2024-07-04T11:42:54Z)
• PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting [59.277480452459315]
  We propose a principled sensitivity pruning score that preserves visual fidelity and foreground details at significantly higher compression ratios. We also propose a multi-round prune-refine pipeline that can be applied to any pretrained 3D-GS model without changing its training pipeline.
  arXiv  Detail & Related papers  (2024-06-14T17:53:55Z)
• Compact 3D Gaussian Representation for Radiance Field [14.729871192785696]
  We propose a learnable mask strategy to reduce the number of 3D Gaussian points without sacrificing performance. We also propose a compact but effective representation of view-dependent color by employing a grid-based neural field. Our work provides a comprehensive framework for 3D scene representation, achieving high performance, fast training, compactness, and real-time rendering.
  arXiv  Detail & Related papers  (2023-11-22T20:31:16Z)
• PRED: Pre-training via Semantic Rendering on LiDAR Point Clouds [18.840000859663153]
  We propose PRED, a novel image-assisted pre-training framework for outdoor point clouds. The main ingredient of our framework is a Birds-Eye-View (BEV) feature map conditioned semantic rendering. We further enhance our model's performance by incorporating point-wise masking with a high mask ratio.
  arXiv  Detail & Related papers  (2023-11-08T07:26:09Z)
• TriVol: Point Cloud Rendering via Triple Volumes [57.305748806545026]
  We present a dense while lightweight 3D representation, named TriVol, that can be combined with NeRF to render photo-realistic images from point clouds. Our framework has excellent generalization ability to render a category of scenes/objects without fine-tuning.
  arXiv  Detail & Related papers  (2023-03-29T06:34:12Z)
• Ponder: Point Cloud Pre-training via Neural Rendering [93.34522605321514]
  We propose a novel approach to self-supervised learning of point cloud representations by differentiable neural encoders. The learned point-cloud can be easily integrated into various downstream tasks, including not only high-level rendering tasks like 3D detection and segmentation, but low-level tasks like 3D reconstruction and image rendering.
  arXiv  Detail & Related papers  (2022-12-31T08:58:39Z)
• EPCL: Frozen CLIP Transformer is An Efficient Point Cloud Encoder [60.52613206271329]
  This paper introduces textbfEfficient textbfPoint textbfCloud textbfLearning (EPCL) for training high-quality point cloud models with a frozen CLIP transformer. Our EPCL connects the 2D and 3D modalities by semantically aligning the image features and point cloud features without paired 2D-3D data.
  arXiv  Detail & Related papers  (2022-12-08T06:27:11Z)
• P2P: Tuning Pre-trained Image Models for Point Cloud Analysis with Point-to-Pixel Prompting [94.11915008006483]
  We propose a novel Point-to-Pixel prompting for point cloud analysis. Our method attains 89.3% accuracy on the hardest setting of ScanObjectNN. Our framework also exhibits very competitive performance on ModelNet classification and ShapeNet Part Code.
  arXiv  Detail & Related papers  (2022-08-04T17:59:03Z)
• Simple and Effective Synthesis of Indoor 3D Scenes [78.95697556834536]
  We study the problem of immersive 3D indoor scenes from one or more images. Our aim is to generate high-resolution images and videos from novel viewpoints. We propose an image-to-image GAN that maps directly from reprojections of incomplete point clouds to full high-resolution RGB-D images.
  arXiv  Detail & Related papers  (2022-04-06T17:54:46Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.